On the Use of Unrealistic Predictions in Hundreds of Papers Evaluating Graph Representations
This addresses a widespread methodological flaw in graph representation research, potentially affecting reproducibility and practical applicability, though it is incremental in proposing fixes rather than new paradigms.
The paper identifies that hundreds of graph representation learning papers use unrealistic predictions by assuming known label counts in multi-label node classification, leading to over-estimated performance, and proposes simple, effective settings without such information while conducting a fair comparison of major methods.
Prediction using the ground truth sounds like an oxymoron in machine learning. However, such an unrealistic setting was used in hundreds, if not thousands of papers in the area of finding graph representations. To evaluate the multi-label problem of node classification by using the obtained representations, many works assume in the prediction stage that the number of labels of each test instance is known. In practice such ground truth information is rarely available, but we point out that such an inappropriate setting is now ubiquitous in this research area. We detailedly investigate why the situation occurs. Our analysis indicates that with unrealistic information, the performance is likely over-estimated. To see why suitable predictions were not used, we identify difficulties in applying some multi-label techniques. For the use in future studies, we propose simple and effective settings without using practically unknown information. Finally, we take this chance to conduct a fair and serious comparison of major graph-representation learning methods on multi-label node classification.